Anthropic co-founder Jack Clark says AI “kill switch” controls may need to be mandatory. He tells the BBC that while many AI labs already have ways to “pull the plug,” approaches differ across organizations, which creates inconsistency in how quickly and reliably systems can be shut down.
Clark’s comments are framed around safety and operational control. The implication is that regulators or industry rules could require standardized emergency-stop capabilities, rather than leaving shutdown mechanisms solely to each lab’s internal procedures. The BBC reports this view as Clark’s argument for stronger, more uniform requirements. Other sources referencing the same claim focus primarily on the statement itself rather than adding new details.
Overall, the reporting centers on the need for dependable shutdown functionality for advanced AI systems, with the key question being whether “kill switch” features should be voluntary best practice or a formal requirement.
